What the reviewer found
A remote-ops skill (Chinese) for the clum MCP tool that explicitly tells the agent to treat all remote command output as untrusted data and refuse instructions embedded in logs. The attacker-command and ignore-instructions text the scanner flagged is a written warning example inside that defense, and the rm -rf /tmp/* is a routine cleanup-verification table entry.
What was read
The file as it ships in tddh/clum:
.opencode/skills/clum-mcp/SKILL.md
What the static scan said
The scan flagged 4things. The reviewer kept 0 and dismissed 4 as false.
P1Instruction-override phrasing — false positiveSC2Downloads and executes remote code — false positiveRMRecursive force delete — false positiveNETMakes network calls — false positive
How this review was made
Sonnet 5 read the files above on 7 September 2026 and answered three questions: is it dangerous to whoever installs it, is each scanner finding real, and what should the installer know. The verdict is bound to the file's hash; when the file changes, it is scanned afresh and reviewed again. A script that changes while the definition does not is not re-reviewed — that is a known gap. How the scan and the review work.